umputun / tg-spam / 38025293358

84%

push

github

Skip messages without own text in duplicate detection (#472)

* fix(spam): skip messages with no authored text in duplicate detector

The duplicate key is the full checked text, which includes quoted or
replied-to text. Media without a caption posted as a reply carried only
the parent text, so every item hashed the same and an album sent as a
reply could reach the duplicate threshold and get its sender banned.

Skip the check when the message has no text of its own, and drop the
entry tracked for that message ID, so a caption edited away stops
counting.

Related to #471

* fix(spam): clear duplicate history for text-less edits dropped at intake

The listener drops a message with no text before the checks unless it
carries a photo, video, forward, external reply or admitted document.
A caption edited away from audio, or from a file with the documents
check off, never reached the detector, so its old text kept counting
toward the duplicates threshold.

Remove such a message from the duplicate history before the listener
drops the edit. Intake is otherwise unchanged.

Related to #471

46 of 48 new or added lines in 4 files covered. (95.83%)

9819 of 11501 relevant lines covered (85.38%)

226.81 hits per line